One of the most successful competitors at this year's Pwn2Own Berlin says powerful AI tools may soon make human hacking contests obsolete.
Key Points:
- A top Pwn2Own Berlin winner says AI models like Claude Mythos could soon outwork even elite hackers.
- Hackers at the contest collectively earned nearly $1.3 million for 47 new vulnerabilities.
- Anthropic restricts Mythos to a small group of governments and security firms over abuse fears.
Pwn2Own Champion Sounds The Alarm
Valentina Palmiotti, a security researcher known as Chompie, was the most successful individual at the annual Pwn2Own contest in Berlin. She told the BBC that she entered the competition thinking it could be her last, a worry rooted in the rapid rise of AI tools built to find software flaws.
That is a striking position from a contest veteran.
Palmiotti won $20,000 on day one for breaking a system tied to Nvidia, then took $50,000 the next day by hacking a Linux based system. She described the overnight grind between the two as exhausting and unhealthy, fueled by energy drinks and adrenaline.
Hackers at the event collectively uncovered 47 brand new methods and earned close to $1.3 million. Every flaw was reported to the affected companies, which are now patching them.
Also Read: XRP Loses Key Support, Now Eyes A Drop Toward $1.31
Why Claude Mythos Worries Researchers
Palmiotti said AI currently sits in a "sweet spot" where it works as an aid, helping her move faster both in competition and in her day job at IBM X-Force. She expects that balance to shift soon.
She pointed to models like Anthropic's Claude Mythos as the turning point. The company says the model is far ahead of any rival in cyber capability, and it has been restricted to a select group of governments and security institutions.
Palmiotti believes good hackers will no longer be needed soon, and that only the very best will keep finding fresh bugs.
Not everyone shares her gloom. Orange Tsai, a Taiwanese hacker who led his team to a $375,000 haul in Berlin, described AI as a capable assistant rather than a replacement. He hopes human intuition will keep spotting flaws that machines miss.
Palmiotti argued the broader trend favors defenders, as long as the strongest tools reach responsible hands first.
The debate arrives as officials weigh the same technology. Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell met Wall Street bank chiefs in April to discuss cybersecurity risks tied to Mythos, a sign of how seriously regulators now treat AI-driven hacking.
Read Next: Ethereum Network Empties Out As Staking Locks A Record 32% Of Supply
